AI RESEARCH
Aligned, Orthogonal or In-conflict: When can we safely optimize Chain-of-Thought?
arXiv CS.AI
•
ArXi:2603.30036v1 Announce Type: cross Chain-of-Thought (CoT) monitoring, in which automated systems monitor the CoT of an LLM, is a promising approach for effectively overseeing AI systems. However, the extent to which a model's CoT helps us oversee the model - the monitorability of the CoT - can be affected by